[FUDZILLA] AMD's initial 28nm lineup (SI) will be a die-shrink only

[MrK6]

We can't say that for certain.

"The folks at DigiTimes say they've gotten word from sources at graphics card makers that the Radeon HD 7000 series, a.k.a. Southern Islands, will hit mass production next month." - TechReport

Wow, that's quick. Isn't the general timeline from starting production to release about 3 months maximum?

 

[digitaldurandal]

Wow, that's quick. Isn't the general timeline from starting production to release about 3 months maximum?

I suppose it could be an indication that the rumor of only a die shrink and no other changes could be true. Without taping out a different architecture or perhaps just completing the original intended architecture - they probably did not need as much time.

 

[load81]

Bah I hope not. I cant resist a new series and was hoping to ride out the 6870/570 for a while. The wife will not be happy.

 

[Idontcare]

Wow, that's quick. Isn't the general timeline from starting production to release about 3 months maximum?

I'm sure you meant minimum, if you did then yes you are correct.

 

[Silverforce11]

SA had stated that AMD taped out 7xxx a month ago. So if they plan mass production next month, it means their tapeout was pretty good.

Would be extremely awesome if 28nm GPUs come in Q4 2011, just in time for my new rig and BF3.

 

[MrK6]

I'm sure you meant minimum, if you did then yes you are correct.

Yep, was thinking maximum time from now until we see cards on the market. If they could squeak it in to 2011, that'd be sweet, although I don't think if I could get one until 2012.

 

[Concillian]

I wouldn't count on lower TDP (power/heat). They will just cram more into the chips and it ends up being just as power hungry (or more).. but obviously perf/watt would go up by a lot.

This depends on your needs.

If you buy with a specific resolution in mind, then yes, you'll certainly be able to get the same or better performance at a lower price and lower TDP.

example: 4850 --> 5770
faster (not a ton, but faster in every way), cheaper, lower TDP.

Of course if you compare a given price point, they'll keep TDP and price similar and it'll be a faster card. But if you're buying for performance at a specific res, it often doesn't make sense to go too much faster. We haven't seen games that are really slowing cards down much recently, we've just seen resolution go up and up and up. Used to be the big res was 1280x1024 or 1600x1200. HD came around and it grew to 1920x1080. Now 2560x1600, but that's only 30 inch monitors? It's growing to the point where you can't get bigger without having to move your head around.

This is why they might drag their feet a bit, It's getting to the point where you need to spend $1000 on display(s) to go with your $500+ video card to even use it's capabilities. The push to higher resolution is shrinking the market by increasing price of display + card. At least until they start significantly increasing actual resolution (dpi) and you can get 2560x1600 on a more sane screen size like 24". As entrenched as the world is with "1080p", I don't see that happening soon though.

 

[3DVagabond]

When we can render photo-realistic animation in real time at eyefinity resolutions in 3D, VGA's will be fast enough. We're not even scratching the surface yet.

 

[biostud]

What features exactly would you like to be added besides more performance/watt/$$$?

CF scaling has become very good with the 6xxx series, and the full range of DX11 features are still to be implemented in games fully.

 

[Silverforce11]

I would like AMD to really push stream support for most apps. So its more software than hardware. As a 28nm die shrink, huge perf/watt/$$ gains will be a given.

 

[AtenRa]

With a shrink in lithography we will gain two things relative to today’s hardware, same performance lower TDP (smaller die size) or same TDP higher performance (Same die size).

So we could roughly have an HD6970 performance at HD6870 die size and TDP or have 50% more performance at HD6970 die size and TDP.

 

[Idontcare]

SA had stated that AMD taped out 7xxx a month ago. So if they plan mass production next month, it means their tapeout was pretty good.

Would be extremely awesome if 28nm GPUs come in Q4 2011, just in time for my new rig and BF3.

Yep, was thinking maximum time from now until we see cards on the market. If they could squeak it in to 2011, that'd be sweet, although I don't think if I could get one until 2012.

If (1) AMD taped out 7xxx a month ago, and (2) it turned out that A0 silicon was production-worthy, and (3) TSMC's ramp-to-production volume capacity was timed to support a Q2 launch of a high-volume 28nm part...then it would be technically feasible for there to be a Q3 (July-Sept) launch of 7xxx.

(but it might still be economically or logistically unfeasible owing to 28nm yields, price/wafer, and supply chain disruptions stemming from the tsunami/earthquake/nuclear reactor crisis in Japan)

If A0 silicon comes back and it is determined that a respin is necessary (or merely justified if they have time in the schedule, 28nm capacity timeline, etc) then you can add another 3 months to the iterative "while" loop, etc.

 

[Topweasel]

I am curious to know what nvidia would of released on 32nm. I think the 580/570 were what the 480/470 were meant to be. But with the issues nv had with 40nm and not wanting to wait any longer to get DX11 cards on the market they released what they had in the 480/470 and spent the next six months getting it right. Then they released the 580/570.

Perhaps Kepler was what they had planned for 32nm and moved it to 28nm.

Its hard to say. Nvidia practically invented the cycle of new core, die shrink, new core. With ATI some times releasing new cores on new processes.

Considering everything. I think the original plan was that the 200 series would turn into the 300 series, but 40nm didn't treat them right and at the high end they couldn't get anything fast enough to displace the GTX200 series. So they continued tweaking that and using an intermediary in 55nm which was probably close enough to not need a complete rework. The 400 series being a new core, and the 500 being 32nm. But once again having to drop it (probably completely). This time they were in luck because it was just that they had an issue with the new process but that it was their supplier suffering and therefore AMD hit the same wall.

 

[tincart]

Various sources keep on telling me not to expect 28nm chips in 2011 so I'm not going to be caught up in this contagion of hope!

 

[OCGuy]

Even having this discussion is AMD's worst nightmare. They aren't even done releasing 6XXX cards, and now just discussing a possible Q3 launch makesthe threat of cannibalization is very real.

It just doesn't make sense for new cards to come out before Xmas.

 

[RussianSensation]

Even having this discussion is AMD's worst nightmare. They aren't even done releasing 6XXX cards, and now just discussing a possible Q3 launch makesthe threat of cannibalization is very real.

It just doesn't make sense for new cards to come out before Xmas.

Doesn't this generally happen though? While HD69xx series is a new architecture, due to 40nm constraints, the performance increase is negligible against the 5870 (released in September 2009). If you think about it, getting 28nm GPUs in Q4 2011 will be 2 years since HD5870 was released. HD6970 is only about 10-20% faster than 5870 outside of 2560x1600 resolutions. Therefore, a huge performance leap is warranted!

And well, you already know that GTX560/570/580 are not a new generation. They are just refreshes of the GTX470/480 cards. It has already been 12 months since Fermi has been released. Therefore, NV should be launching Kepler within 12 months as well. The only limitation at this point is the maturity of 28nm process. The designs for most camps are most likely ready.

 

[tijag]

They could release a 28nm thats just a die shrink of Cayman and sell it as 6975 or something like that.

Or they could release a die shrunk Cayman, and sell it as a 7770/7750.

 

[MrK6]

If (1) AMD taped out 7xxx a month ago, and (2) it turned out that A0 silicon was production-worthy, and (3) TSMC's ramp-to-production volume capacity was timed to support a Q2 launch of a high-volume 28nm part...then it would be technically feasible for there to be a Q3 (July-Sept) launch of 7xxx.

(but it might still be economically or logistically unfeasible owing to 28nm yields, price/wafer, and supply chain disruptions stemming from the tsunami/earthquake/nuclear reactor crisis in Japan)

If A0 silicon comes back and it is determined that a respin is necessary (or merely justified if they have time in the schedule, 28nm capacity timeline, etc) then you can add another 3 months to the iterative "while" loop, etc.

Great info, thanks! :thumbsup:

Even having this discussion is AMD's worst nightmare. They aren't even done releasing 6XXX cards, and now just discussing a possible Q3 launch makesthe threat of cannibalization is very real.

It just doesn't make sense for new cards to come out before Xmas.

Possibly, but only if they have 6xxx cards to sell. If they stopped production early because the 7xxx cards were done and that much better/cheaper to produce, it might be more than worth it.